[ExI] Let's stop lowering our IQs.

[Anders Sandberg]

On 05/02/2013 10:09, BillK wrote:
> <http://www.evidence.nhs.uk/> is a search engine that includes 
> Cochrane reviews. But that is only for medical and health information.
I was looking for that for my post, thanks for finding it!

> The problem of good searching of reliable sources still exists for
> other domains. 'How to search' is an education in itself,

Yes. It is an important life skill right up there with reading, typing 
and Bayesian inference. And like the others it needs to be trained. Do 
we know any good training programs (for kids or adults) in searching?

I often bring up the importance of connoisseurship - you might not be 
able to do science or carpentry, but you can at least tell whether 
something is a good paper or a good table. Part of why we should acquire 
a wide education is in order to become connoisseurs of many things, so 
we can act as filters for ourselves and friends.

> especially as Google (and
> others) are now editing their search results by secretive methods.
> They claim the secrecy is for good reasons, to avoid people gaming
> their search results. But the impression is that Google themselves may
> be 'gaming' their search results to assist advertisers and downgrade
> sites that they don't like.

Bah, that is just typical human conspiracy thinking (looking for selfish 
intentions and then thinking them to be paramount) obscuring a much 
bigger problem: there is going to be bias because of the basic search 
system. Search systems that handle language or different sources 
differently will give you far more subtle and insidious biases. A search 
system where ranking is roughly proportional to popularity will show you 
what a mainstream thinks, while one based on other network properties 
will show you biases that we do not even have a name for (imagine 
searching using pagerank based on the second or third eigenvector of the 
web adjacency graph instead of the first - there are parallel web 
universes out there!) Good search skills take the mechanical and human 
biases into account.

In a world where Google just ran a straight pagerank and everybody else 
gamed them, you would have exactly the same problem with hidden biases 
since now you would not know what the SEO people were doing.

> And readers being educated enough to be able to read and understand
> scientific papers is still a barrier to understanding.

Not to mention that most papers are boring and badly written. Part of it 
is the fault of science education, which doesn't drill us in academic 
language skills (or the enjoyment of writing). But part of the blame is 
the cultural consensus of how a paper "must" be written, which produces 
the endless row of IMRAD-organised papers in passive third person voice 
where p-values are mentioned but not effect sizes.

-- 
Anders Sandberg,
Future of Humanity Institute
Philosophy Faculty of Oxford University